search

zhaoyuzhi / A-New-Convolutional-Neural-Network-Based-Data-Driven-Fault-Diagnosis-Method

TensorFlow implementation of a CNN based mechanical science paper
45 stars 17 forks source link

data process #1

Open enjoyglance opened 5 years ago

enjoyglance commented 5 years ago

HEY ,I don't know how to process the data ,could you give me the code or instruction how to deal with it ? a little fee is okay! please contact my email :guanheyvgmail@.com or my fb account :Hury Gwan

enjoyglance commented 5 years ago

my wechat:754354719

zhaoyuzhi commented 5 years ago

I think it is better for you to check with the paper.

zhaobeile commented 5 years ago

嘿，我不知道如何处理数据，你能给我代码或指令如何处理它？一点费用还可以！请联系我的电子邮件：guanheyvgmail @ .com或我的fb帐户：Hur

zhaobeile commented 5 years ago

brother,i cann’t get so many picture from the singal ,could you give me the code or instruction how to deal with it ? 1714178171@qq.com

zhaoyuzhi commented 5 years ago

According to the paper Section III: PROPOSED CNN FOR FAULT DIAGNOSIS, . In order to obtain an M × M size image, a segment signal with the length M^2 would be randomly obtained from the raw signal. The mapping function is shown in equation (1) in paper. You may write a simple python script to do that.

zhaoyuzhi commented 5 years ago

brother,i cann’t get so many picture from the singal ,could you give me the code or instruction how to deal with it ? 1714178171@qq.com

The single is 1-D data. You may follow the equation (1) to transfer it into 2-D, and use OpenCV / PIL to save the image.

zhaobeile commented 5 years ago

您好，十分感谢您的回复，我已经尝试过生成图片，将一维数据转化成二维，但是，根据一个mat文件下的数据大概有480000左右，生成的图片也就100左右，我想请教的是如何得到训练集里那么多的图片，是通过滑动平移，每张图片有信号重叠吗？

------------------ 原始邮件 ------------------ 发件人: "Yuzhi ZHAO"notifications@github.com; 发送时间: 2019年7月21日(星期天) 下午2:50 收件人: "zhaoyuzhi/A-New-Convolutional-Neural-Network-Based-Data-Driven-Fault-Diagnosis-Method"A-New-Convolutional-Neural-Network-Based-Data-Driven-Fault-Diagnosis-Method@noreply.github.com; 抄送: "赵威同学"1714178171@qq.com;"Comment"comment@noreply.github.com; 主题: Re:[zhaoyuzhi/A-New-Convolutional-Neural-Network-Based-Data-Driven-Fault-Diagnosis-Method]data process (#1)

brother,i cann’t get so many picture from the singal ,could you give me the code or instruction how to deal with it ? 1714178171@qq.com

The single is 1-D data. You may follow the equation (1) to transfer it into 2-D, and use OpenCV / PIL to save the image.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

zhaoyuzhi commented 5 years ago

Of course it can be overlapped, you'd better to check the paper. Every time you get an image, do not forget to save label (whether it is damaged) also.

zhaobeile commented 5 years ago

当然它可以重叠，你最好检查一下纸张。每次获得图像时，不要忘记保存标签（是否已损坏）。 您是怎么做的，滑动多少个信号开始采集一张图像，如果不重叠，获得的图像是很少的，论文里并没有细说如何采集的，只讲了4096个数据作为一张图片